intelligent client-server applications intelligent agents for e-commerce

Transcription

1 topics: software agents cis20.1 design and implementation of software applications I fall 2007 lecture # III.4: intelligent client-server applications gathering information and using it wisely notes courtesy of Peter McBurney, Univ of Liverpool, UK intelligent client-server applications software agents: also known as intelligent agents, softbots or just bots there are many definitions of agents suffice it to say that an agent is a computer program that responds to inputs from its environment by acting intelligently a robot is an example of an embodied agent but there are many examples of virtual agents on the internet we ll explore some here, to give you an example of how a client-server application can be intelligent 1 2 intelligent agents for e-commerce why? information overload in 2001, 10 billion non-spam messages were sent per day in 2005, this was up to 35 billion messages the amount of unique information produced world-wide is doubling every year the amount of information that was accessible on the internet in 2002 is more than all the information that was accessible in the 300,000 years of human history most data we gather goes un-used! for example, according to the Gartner Group (Kyte 2002): the amount of data collected by large enterprises doubles every year knowledge workers can analyze only about 5% of the data most of knowledge workers efforts (60% or more) are spent trying to discover important patterns in the data a much smaller percentage (20% or more) is spent determining what those patterns mean and very little time (10% or less) is spent actually doing something based on those patterns (!) information overload reduces knowledge workers decision-making capabilities by 50% 3 4

2 what is the solution? value of software agents delegate, do not navigate What we call agent-based interfaces will emerge as the dominant means by which computers and people will talk to one another (Being Digital, 1995) the rise of information filtering: end users delegate tasks such as information review and filtering to mobile agents that travel to remote data sources, examine them locally and return with a summary of the data information access and navigation are the major applications of the software agents in today s internet other tasks: mundane personal activity (e.g., grocery shopping, travel planning) search and retrieval repetitive office activities decision support domain expertise 5 6 definition and concepts comparison An agent is anything that can be viewed as perceiving its environment through sensors and acting on that environment through effectors. (Russell and Norvig 1995, p33) intelligent software agents continuously perform three functions: perception of dynamic conditions in their environment, action to affect conditions in their environment, and reasoning to interpret perceptions, solve problems, draw inferences and determine actions 7 8

3 intelligence levels frequently referred to as semi-intelligent or software agents. Level 3. Agents at this level have a learning and deductive component of user profiles to help a user who cannot formalize a query or specify a target for a search. DiffAgent (CMU) and Letizia (MIT) are examples of such agents. Agents at this level are referred to as learning or truly intelligent agents. Similar to the concept of levels is the idea of agent generation definitions of agents are greatly dependent on the agents levels of intelligence, which are described by Lee et al.(2002) as follows: Level 0 (the lowest). These agents retrieve documents for a user under straight orders. Popular Web browsers fall into this category. The user must specify the URLs where the documents are. These agents help in navigating the Web. Level 1. These agents provide a user-initiated searching facility for finding relevant Web pages. Internet search agents such as Google and Ask.com. Information about pages, titles, and word frequency is stored and indexed. When the user provides key words, the search engine matches them against the indexed information. These agents are referred to as search engines. Level 2. These agents maintain users profiles. They then monitor the Internet and notify the users whenever relevant information is found. An example of such an agent is WebWatcher (search for WebWatcher at cs.cmu.edu). Agents at this level are

4 shopbots a classification of bots shopbots are computer programs that visit web sites also called: agents, spiders, robots, softbots, bots the activity of collecting information from web pages is often called harvesting example applications: to find a particular product: e.g., to find a rare book or CD to comparison shop: e.g., to find all the sites selling a particular computer game and find the cheapest e.g., to find all the sites selling a particular computer game and find the one that will deliver it the fastest to gather information e.g., for web-page indexing; to gather addresses; to archive publications 13 chatterbots a bot that undertakes a conversation with a human e.g., a front-end to a web-search engine the first and most historically famous chatterbot was Eliza (Weizenbaum 1966); she gave psychotherapy to humans in natural language commercial intermediaries bots that act in between two or more companies e.g., brokering deals, monitoring work (e.g., delivery of goods), etc. government bots helping citizens obtain government information or fulfill legal obligations e.g., helping with car licensing or new business registration processes news bots e.g., retrieving news stories; creating and delivering personalized e-newspapers 14 newsgroup bots help with management of newsgroups e.g., screening abusive/obscene language; creating group postings, etc. update bots alerts users about changes e.g., to a web site; about software updates; about virus protection web-developer bots e.g., check if links are still current; track hits shopbots shopping bots bots that act on behalf of human users who are potential customers for some product or service classification due to Patti Maes (MIT Media Lab) product brokering bots alert users to new releases of products bots recommend products based on past purchases or stated user preferences usually works with one potential supplier people who bought X also bought..., people who liked Z also liked... examples: matching is a big and open research problem check out the Netflix prize ( merchant brokering 15 16

5 bot undertakes comparison shopping across a range of potential suppliers, collecting information, collating and analyzing it, and presenting it to the user works with multiple suppliers negotiation bot enters auctions or undertakes negotiations on behalf of user, within parameters set by user counter-parties (i.e., those who the agents represent) may be humans or other bots e.g., electronic traders design decisions for shopbots what is the human-computer interface? e.g., users can be overwhelmed if too much information is presented to them, if too many questions are asked about preferences how are preferences elicited from users? it is typically difficult for users to know a priori or not in situ what their preferences are which potential sellers to include are potential sellers hard-wired (known a priori) or is the search done on-the-fly (dynamically)? most are hard-wired because: resulting searches are faster sellers have to provide information in an agreed-upon format e.g., Amazon API ( but on-the-fly searches can produce new, up-to-the-minute information... when does the shopbot stop waiting to receive an offer? how are offers short-listed? they may use some model of multi-attribute decision-making how are offers displayed display of information can influence user s choice! how much autonomy (ability to make its own decisions) does the shopbot have? can the shopbot learn from experience? e.g., to learn the user s preferences by observing his/her actual purchase decisions e.g., to learn which suppliers are reliable can the shopbot predict market trends? e.g., to suggest to the human that prices may fall/rise if s/he waits... what aspects of the human decision are included? e.g., cost of waiting? cost of information processing? what about XML? many web services communicate with each other using XML do all vendor sites need to use XML? no because most sites have certain regularities navigation regularities (so users can find stuff quickly) uniform look-and-feel vertical separation (new products on a new page/separate part of a page) common symbols (e.g., prices quoted with $ in front of them) difficulties arise in some domains many are not standard (e.g., entertainment industry sites) some use text embedded in graphics hard or impossible to parse common for company logos increasingly common for addresses and other contact details to cut down on spam 19 20

6 parsing may need to be vendor-specific XML has conventions, but it is a template language contents are not specified as part of the language semantic web tool that provides a content presentation and organization standard so that content can be shared safely among different software applications, using mutually understandable semantic constructs hot research area! current standards for web services agents as web services XML (extensible markup language) provides the common service providers and requestors need to connect and exchange information SOAP (The simple object protocol) provides the common protocol systems need to communicate with each other so that they can request services. For example: schedule appointments, order parts and deliver information WSDL (the web services description language) describes the services in a machine readable form, where names of functions, their required parameters, and their results can be specified UDDI (universal descriptions, discovery, and integration) gives clients (individuals and businesses) a way to find needed services by specifying a registry or yellow pages of services 23 24

7 advantages of semantic web limitations of semantic web easy to understand semantic Web demonstrates objects and their relationships as graphic templates for easy understanding easy resource integration easier to integrate systems and modules designed in semantic web; this also makes it easier for system analysis and maintenance saving development time and costs semantic Web allows incremental ontology creation, enabling more rapid system development and lower development costs. automatic update of content because agents can easily locate a specific knowledge on semantic web, they can have functions to update or import contents automatically; this adds the level of intelligence to the software agents easy resource reuse ontology-based annotations can turn briefings into reusable resources 25 graphical representation may be oversimplified for example, using an arrow to represent a relation between two instances is unable to show more complicated multiparty relations additional tools for searching content and building references to pre-existing instances are needed for effective use of semantic webs ontologies may not be correctly defined in some cases, the outcome of this could be severe; it is still hard to prove the completeness or correctness of a defined ontology unreliability when agents deal with a semantic web containing information that is inconsistent, incorrect, or unreliable, the agents could become contaminated or be misled security because the semantic web allows agents from different systems to communicate and share information, security is a key concern. Security is always a problem for an open system. 26